Legal AI is moving beyond drafting assistance toward coordinated, multi-step workflows. In a case study published on October 28, 2026, OpenAI profiles Legora, the Stockholm-founded legal technology company building what its CEO calls an “agentic operating system for legal work.” The company uses OpenAI models to power legal analysis and agent-driven document review, while ChatGPT and Codex also support its internal teams.
The most attention-grabbing results are substantial: Legora reports reducing the share of its internal legal team's capacity spent reviewing nondisclosure agreements from roughly 20% to less than 1%; one Fortune 100 customer reports 90% faster contract turnaround, more than $20 million in external litigation and diligence savings, and 9.5 hours returned per lawyer each week.
These figures come from Legora and a customer example presented in OpenAI's case study. They are not independent, cross-industry benchmarks and should not be interpreted as guaranteed savings for every law firm or legal department.
From legal AI assistant to agentic operating system
Legora was founded in Stockholm in 2023. Its CEO and co-founder Max Junestrand describes the company's ambition as making legal intelligence available on demand. The problem is familiar to legal teams: contracts, due diligence materials and routine documentation often exceed the time available for thorough human review.
Earlier generations of legal software primarily organized documents, supported searches or automated narrow tasks. Legora's approach uses increasingly capable AI models and agents to analyze collections of documents, apply structured questions and surface issues for legal professionals to evaluate. The aim is not simply to write faster; it is to increase the breadth of work that a legal team can realistically complete.
OpenAI says its models are a key technology provider for Legora's platform. Legora uses them in core workflows spanning agentic legal work and large-scale document review, incorporating new capabilities as the underlying models evolve.
What “agentic” means in this context
An AI agent differs from a basic text generator when it can pursue a defined objective through multiple steps, use tools and manage intermediate results. In legal workflows, this may involve identifying relevant contracts, checking clauses against a playbook, organizing findings and preparing exceptions for professional review.
The word agentic should not be mistaken for unrestricted autonomy. The case study emphasizes that lawyers continue to review, interpret and make consequential decisions. Legal advice, negotiation strategy, privilege and client obligations cannot be reduced to an automated checklist.
Legora's own NDA workflow: from 20% of capacity to less than 1%
Legora's General Counsel, Emma Kadri Bergström, describes how her team examined previously signed nondisclosure agreements and created a playbook specifying acceptable terms. The system can then conduct a first pass on incoming NDAs and highlight clauses that fall outside those parameters.
Before the process was introduced, Bergström estimates NDA review consumed about 20% of the team's capacity. Today, it accounts for less than 1%, which OpenAI summarizes as a reduction exceeding 95%.
That is a striking operational improvement, but the denominator matters. The figure describes the proportion of one company's legal-team capacity allocated to NDA review; it does not mean every NDA is processed 95% faster or that the system is accurate 95% of the time.
Legora's internal legal team has also had to support rapid organizational growth. According to the case study, the business expanded from three offices to 18 over the prior year, completed three financing rounds and made five acquisitions. Recovering capacity from repeatable work can help a team handle complex transactions without treating every new request as a staffing emergency.
Due diligence: the opportunity is broader review, not just faster review
Large transactions may involve thousands of contracts and numerous questions about liability, assignment rights, termination clauses or commercial obligations. Traditionally, deadlines and budgets can force lawyers to focus detailed attention on a subset of the documents.
Alex Fortescue-Webb, Legora's Global Head of Legal Engineering, describes how its Tabular Review workflow can analyze thousands of contracts across multiple questions in minutes. That creates the possibility of examining a broader document set and surfacing issues that might otherwise be missed because the relevant contract was outside the manually prioritized sample.
Crucially, an AI system's ability to process more pages is not the same as proof that every conclusion is correct. The useful output is a structured first pass that lawyers can verify, challenge and contextualize.
The Fortune 100 results: what the numbers actually show
OpenAI's October 28 case study highlights one unnamed Fortune 100 Legora customer with three reported outcomes: 90% faster contract turnaround, more than $20 million saved on external litigation and diligence, and 9.5 hours returned per lawyer per week.
These metrics describe one customer's experience as reported by Legora. The published article does not provide a full independent methodology, control group or a detailed breakdown of the savings calculation. Readers should therefore treat them as a customer case study rather than an industry-wide forecast.
Nevertheless, the example illustrates two distinct sources of value: reducing the elapsed time needed to complete work and increasing the volume or depth of analysis that teams can undertake within the same staffing constraints.
How ChatGPT, Codex and coding agents affect Legora's own operations
Legora is not only selling AI to legal teams. OpenAI says the company has expanded use of ChatGPT Work, Codex and dots across its internal operations. The case study gives particularly concrete examples from engineering.
CTO Jacob Lauritzen says coding agents allow his teams to prototype many approaches in parallel rather than commit early to a single engineering solution. He describes exploring as many as 20 candidate solutions at once and selecting a promising option within days.
In incident response, coding agents can inspect millions of lines of logs and investigate possible causes simultaneously. In some cases, Legora reports going from a bug report to a working fix in about five minutes. That is a company example of an accelerated troubleshooting process, not a guaranteed response time for all incidents.
For customers, these internal improvements may translate into quicker bug fixes, faster iteration and earlier feedback loops. For software businesses more broadly, the implication is that agentic tools can affect both the product delivered to users and the speed at which the product itself is built.
Another Legora test: financial-statement review with GPT-6 Astra
A separate OpenAI case study published September 3, 2026 offers a more specific technical example. In a financial-statement tie-out workflow, Legora says an agent using GPT-6 Astra reviewed 41 documents in minutes, found all four deliberately planted errors and improved performance by nearly 40% against the previous model on that particular task.
The same report notes that the average improvement across Legora's broader Benchmark for Agentic Reasoning was approximately 3%. That contrast is important: a large gain on one workflow should not be presented as the model's universal improvement across legal tasks.
The test illustrates how context processing, structured checking and traceable outputs can support document-heavy work. It also reinforces the need for domain-specific evaluation rather than relying on generic model capability claims.
Why human legal judgment remains essential
Contracts are not just collections of clauses. Their meaning depends on jurisdiction, factual circumstances, negotiating history, commercial objectives and the relationship between provisions. A system may correctly identify a clause while misunderstanding its practical implications.
Legal AI deployments therefore need robust review procedures, documented sources, access controls, confidentiality protections and escalation rules for uncertain findings. Organizations should determine when a lawyer must approve an action, how evidence is preserved and how incorrect outputs can be corrected.
These safeguards are especially important in litigation, due diligence and regulated transactions. Efficiency gains are valuable only if they do not compromise client confidentiality, legal privilege or professional responsibility.
Five lessons for enterprise AI adoption
OpenAI's Legora profile also presents broader operational lessons. First, leadership use matters: Legora reports that when more than 75% of leaders actively use the platform, median organizational usage is three times higher. This is an observed association in its reported data, not proof that leadership use alone causes adoption.
Second, organizations should begin with a specific workflow and measurable problem, rather than deploying agents without a clear purpose. Third, internal champions can help turn experiments into repeatable practices. Fourth, teams should expect to revisit processes as model capabilities change. Fifth, evaluation, transparency and user trust need to be designed into the workflow from the outset.
These principles apply well beyond law, including financial operations, compliance, professional services and enterprise content production.
What this means for SEO and digital agencies
Legora's legal workflows are not SEO ranking technology, and the case study does not announce any change to Google Search or AI Search visibility. Its relevance to digital agencies is organizational: agentic systems may allow specialists to supervise larger volumes of research, auditing, reporting and quality assurance without treating automated output as final authority.
For example, an SEO agency could use agents to identify technical issues across many websites, organize evidence and propose priorities. But human specialists should still validate crawl data, check the business impact, review recommendations and approve consequential changes.
Like legal teams, agencies should measure both time saved and the quality of decisions. Faster output is not automatically better output, and scaling a flawed process with AI can multiply errors.
The next frontier: overnight analysis and real-time collaboration
Legora's CTO describes two possible future directions. One is longer-running agents that could review extremely large document collections overnight, leaving organized findings ready for a lawyer in the morning. The other is faster, more interactive collaboration in which an AI system and a professional work back and forth on a document or question in real time.
These are ambitions and forward-looking scenarios, not capabilities demonstrated at the scale of millions of pages in the October case study. As organizations move toward longer-running agents, auditability and human review will become even more important.
NetContentSEO assessment
The most important aspect of Legora's story is the shift from isolated AI assistance to integrated professional workflows. The company reports measurable time savings, wider document analysis and faster software development, while preserving a central role for legal expertise.
For businesses evaluating agentic AI, the lesson is to start with bounded tasks, build clear evaluation methods and expand only when outputs are reliable enough for the intended risk level. Legora's results are promising, but they should be understood as company-reported examples rather than a promise of universal productivity gains.
The future of professional work may involve fewer repetitive steps and more time spent on interpretation, client relationships and strategic decisions. Achieving that outcome depends not only on better models, but also on how organizations design accountability around them.
Primary sources: OpenAI, “Legora is building the agentic operating system for legal work,” October 28, 2026; OpenAI, “Legora reviewed 41 documents in minutes with GPT-6 Astra,” September 3, 2026.